Agents are in Confluence (and wherever you need them to be)

Atlassian has extended AI agents in Confluence to perform tasks such as commenting, labeling, setting statuses, and creating whiteboards or databases, mirroring human teammate actions. These agents can be summoned directly within the editor or comments and operate within predefined permissions, ensuring they only act where authorized. Since their May 2024 launch, agents have executed over 5 million monthly invocations, saving customers more than 200,000 hours in February alone. The agents are now accessible not just in Confluence but also through the Atlassian Rovo MCP in tools like Claude, Cursor, ChatGPT, or integrated development environments (IDEs).

Agents now handle routine tasks such as processing feedback, drafting status updates, or preparing decision materials, allowing teams to focus on higher-value work. For example, Riverty’s Lessons Learned agent mines past projects to prevent repeated mistakes, while KFC’s Architecture Review agent checks proposals against document standards before review. Pythian’s Progress Tracker aggregates Jira and Confluence data to draft status emails, and Sprout Social’s Onboarding agent addresses 80% of new-hire questions automatically. These agents operate without requiring custom code, as users can describe tasks in plain English and build them in Rovo Studio for team-wide use.

The agents function on the Teamwork Graph, Atlassian’s context layer that links work, people, knowledge, and code across Atlassian products and connected tools. According to Atlassian’s benchmarking, agents using Teamwork Graph context delivered 44% more accurate results while reducing token usage by 48%. Agentic workflows in Confluence are supported through Rovo, available on Standard, Premium, and Enterprise Cloud plans, with all Atlassian Cloud customers able to access the Atlassian Rovo MCP Server. MCP actions currently run on the preview endpoint until MCP 2.0 is released.

The expansion follows the introduction of the Atlassian Rovo MCP, which provides coding agents with deeper, scoped access to complete tasks in Jira while maintaining critical context directly in IDEs or terminals. This update introduces a new class of agent skills designed to enhance teamwork and integration with everyday applications, aiming to streamline workflows and reduce manual effort across development and collaboration tools.